Fill in this form to receive a download link:
- Interactions with artificial agents often lack immediacy because agents respond slower
- 20 than their users expect. Automatic speech recognisers introduce this delay by analysing a
- 21 user’s utterance only after it has been completed. Early, uncertain hypotheses of incremental
- 22 speech recognisers can enable artificial agents to respond more timely. However, these
- 23 hypotheses may change significantly with each update. Therefore, an already initiated action
- 24 may turn into an error and invoke error cost. We investigated whether humans would use
- 25 uncertain hypotheses for planning ahead and/or initiating their response. We designed a
- 26 Ghost-in-the-Machine study in a bar scenario. A human participant controlled a bartending
- 27 robot and perceived the scene only through its recognisers. The results showed that
- 28 participants used uncertain hypotheses for selecting the best matching action. This is
- 29 comparable to computing the utility of dialogue moves. Participants evaluated the available
- 30 evidence and the error cost of their actions prior to initiating them. If the error cost was low,
- 31 the participants initiated their response with only suggestive evidence. Otherwise, they waited
- 32 for additional, more confident hypotheses if they still had time to do so. If there was time
- 33 pressure but only little evidence, participants grounded their understanding with echo
- 34 questions. These findings contribute to a psychologically plausible policy for human-robot
- 35 interaction that enables artificial agents to respond more timely and socially appropriately
- 36 under uncertainty.